A Parallel Framework for Constraint-Based Bayesian Network Learning via Markov Blanket Discovery

Srivastava, Ankit; Chockalingam, Sriram P.; Aluru, Srinivas

doi:10.1109/tpds.2023.3244135

Cited by 7 publications

(2 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Constraint-based algorithms use a series of conditional hypothesis tests to learn independences among the variables in the model. Grow-Shrink (GS) (Srivastava et al ., 2023), PC algorithm (Tsagris, 2019) and Inter-IAMB (Guo et al ., 2022) are widely used in the practice of BN structure learning and have achieved good performance. The score-based algorithms is to find the best DAG according to some score function that measures its fitness to the data (Heckerman et al ., 1995).…”

Section: Literature Reviewmentioning

confidence: 99%

A decision support method for credit risk based on the dynamic Bayesian network

Lu,

Wu,

Dong

et al. 2023

IMDS

View full text Add to dashboard Cite

PurposeCredit risk evaluation is a crucial task for banks and non-bank financial institutions to support decision-making on granting loans. Most of the current credit risk methods rely solely on expert knowledge or large amounts of data, which causes some problems like variable interactions hard to be identified, models lack interpretability, etc. To address these issues, the authors propose a new approach.Design/methodology/approachFirst, the authors improve interpretive structural model (ISM) to better capture and utilize expert knowledge, then combine expert knowledge with big data and the proposed fuzzy interpretive structural model (FISM) and K2 are used for expert knowledge acquisition and big data learning, respectively. The Bayesian network (BN) obtained is used for forward inference and backward inference. Data from Lending Club demonstrates the effectiveness of the proposed model.FindingsCompared with the mainstream risk evaluation methods, the authors’ approach not only has higher accuracy and better presents the interaction between risk variables but also provide decision-makers with the best possible interventions in advance to avoid defaults in the financial field. The credit risk assessment framework based on the proposed method can serve as an effective tool for relevant policymakers.Originality/valueThe authors propose a novel credit risk evaluation approach, namely FISM-K2. It is a decision support method that can improve the ability of decision makers to predict risks and intervene in advance. As an attempt to combine expert knowledge and big data, the authors’ work enriches the research on financial risk.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

A decision support method for credit risk based on the dynamic Bayesian network

Lu,

Wu,

Dong

et al. 2023

IMDS

View full text Add to dashboard Cite

show abstract

“…Learning high-quality structural models from sample data is the key to solving practical problems with BN theory. Accurate calculation of BN structure learning is an NP-hard problem [13]. Therefore, some scholars have proposed using heuristic algorithms to solve this problem.…”

Section: Introductionmentioning

confidence: 99%

Self-Awakened Particle Swarm Optimization BN Structure Learning Algorithm Based on Search Space Constraint

Liu,

Li,

Zhang

et al. 2023

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

To obtain the optimal Bayesian network (BN) structure, researchers often use the hybrid learning algorithm that combines the constraint-based (CB) method and the score-and-search (SS) method. This hybrid method has the problem that the search efficiency could be improved due to the ample search space. The search process quickly falls into the local optimal solution, unable to obtain the global optimal. Based on this, the Particle Swarm Optimization (PSO) algorithm based on the search space constraint process is proposed. In the first stage, the method uses dynamic adjustment factors to constrain the structure search space and enrich the diversity of the initial particles. In the second stage, the update mechanism is redefined, so that each step of the update process is consistent with the current structure which forms a one-to-one correspondence. At the same time, the "self-awakened" mechanism is added to prevent precocious particles from being part of the best. After the fitness value of the particle converges prematurely, the activation operation makes the particles jump out of the local optimal values to prevent the algorithm from converging too quickly into the local optimum. Finally, the standard network dataset was compared with other algorithms. The experimental results showed that the algorithm could find the optimal solution at a small number of iterations and a more accurate network structure to verify the algorithm's effectiveness.

show abstract